Regression: impossible to dump multiline strings with leading newline

Pull request #551 changed the behaviour of raw, non-escaped string values. Running

import tomlkit
s = tomlkit.string(
  "\n  aligned\n  strings\n",
  multiline=True,
  literal=True,
  escape=False,
)
print(tomlkit.dumps(s))

used to output

'''
  aligned
  strings
'''

as expected, but now outputs

'''

  aligned
  strings
'''

(note the extra newline). In fact, it is now impossible to produce the desired output.

Proposed fix: hide the behaviour change of #551 under the escape flag.

#658 is a clean fix that does not change the API.

But the situation is not ideal because under escape=False the value of the string does not match the parsing of the original.

More principled solutions that would change the API:

Have the string function take two optional string arguments value and original, one of which must be given. If original is None, it is computed from value as currently with escape=True. If value is None, it is parsed from original. If both are given, one might have a sanity check that the parsing of original agrees with value.
Have two separate String building functions, one taking a parsed string and one taking an unparsed string.

Navigation Menu